EU AI Act labeling requirements: what an AI creator has to do
Since 2 August 2026, Article 50(4) of the EU AI Act requires anyone who uses AI professionally to make deepfake images, audio or video to disclose that the content is AI-generated, and the Commission’s FAQ counts regular income from it as professional. For a monetised AI influencer, that very likely means a visible AI label on her content from the first time anyone sees it. This page goes deeper on the EU row of our AI influencer rules.
On this page
- What does the EU AI Act require you to label?
- Are you a deployer if you run an AI influencer?
- Is a fictional AI character a “deepfake”?
- Does the EU AI Act apply if you’re outside the EU?
- How do you label AI content under Article 50?
- What are the EU’s AI labelling icons?
- What is the Code of Practice on AI labelling?
- What happens on 2 December 2026?
- What are the fines for breaking Article 50?
- What can go wrong?
What does the EU AI Act require you to label?
Article 50 splits the job in two. Tool makers (“providers”) must mark AI outputs in a machine-readable way. People who use the tools professionally (“deployers”) must disclose deepfakes in a way people can see. A creator publishing photorealistic AI images is on the deployer side. Both duties apply from 2 August 2026.
From the AI Act text, checked 2 October 2026:
| Article | Who | What it requires | From |
|---|---|---|---|
| 50(2) | Providers of generative AI | Mark outputs in a machine-readable format, detectable as AI | 2 Aug 2026 (2 Dec 2026 for systems already on the market) |
| 50(4) | Deployers | Disclose that deepfake image, audio or video content is AI-generated or manipulated | 2 Aug 2026 |
| 50(4) | Deployers, artistic or fictional work | Disclose “in an appropriate manner that does not hamper the display or enjoyment of the work” | 2 Aug 2026 |
| 50(5) | Both | Information must be “clear and distinguishable”, at the latest at first interaction or exposure | 2 Aug 2026 |
Are you a deployer if you run an AI influencer?
If she earns money, very likely yes. The AI Act excludes use “in the course of a personal non-professional activity”. The Commission’s FAQ says a person who gains “an economic benefit on a regular basis” from the activity is doing it professionally, which makes them a deployer. Fanvue subscriptions, UGC fees and paid posts would almost certainly count.
The Commission’s FAQ on Article 50 (last updated 24 July 2026) gives the contrast directly. Someone who makes deepfakes for social media in a personal capacity is outside the Act. The same person is a deployer once it’s a business, trade, occupational or freelance activity. An AI influencer built to earn on Fanvue or from brands is the second case.
Is a fictional AI character a “deepfake”?
Probably, when she’s photorealistic. The Act defines a deepfake as AI content that “resembles existing persons, objects, places, entities or events” and would falsely appear authentic. The Commission’s FAQ reads “existing” broadly: things that exist “can plausibly exist or could have plausibly existed”. A realistic woman who could plausibly be real fits that.
The FAQ lists three criteria that must all be met:
- Resemblance: a high level of similarity to the subject.
- Existing: the subject resembles someone or something that exists or could plausibly exist.
- False appearance: the content could mislead someone into thinking it’s authentic.
Whether her content meets the third one depends on context, and that’s an interpretation question we can’t settle for you. The practical answer doesn’t change: label her anyway. Disclosure costs nothing, and the platforms require it regardless. If her content counts as “evidently” fictional or creative work, the duty is lighter, not gone: you still disclose, just in a way that doesn’t spoil the work.
Does the EU AI Act apply if you’re outside the EU?
It can. Article 2(1)(c) applies the Act to providers and deployers in a third country “where the output produced by the AI system is used in the Union”. An account with followers in the EU publishes output that’s used there. If any of your audience is in the EU, assume Article 50 applies to you.
It also overlaps with what the platforms ask for anyway. In our reading, Instagram’s AI-generated profile label, TikTok’s AI-generated content toggle and Fanvue’s disclosure rule cover most of what Article 50 asks of a creator. Doing those properly is most of the work.
How do you label AI content under Article 50?
Visibly, from the first time someone sees it. The Commission’s FAQ says disclosure must be “understandable and perceivable by natural persons”, without special tools, and that deployers “cannot simply rely on the machine-readable marking” a tool embeds. Metadata in the file doesn’t count as your label. You need something a person can see.
A label set that covers this on every platform the course uses:
- Profile level: turn on Instagram’s AI-generated profile label on every account that features her, and put “AI-generated” in the bio on Instagram, TikTok, Threads and Fanvue.
- Post level: switch on the AI label on each post. Instagram’s version is its AI info label, explained alongside the profile label; TikTok’s is the AI-generated content toggle.
- In the image, optional: place the EU’s “Fully AI-Generated” icon in a corner, or a small “AI” watermark. On Fanvue a watermark is one of the accepted disclosure methods.
- Captions: a short “AI-generated” line on anything that could be reshared without the platform label.
- Keep the metadata. Don’t strip marks the tool embedded. They don’t replace your label, but removing them works against it.
What are the EU’s AI labelling icons?
A free set of icons from the Commission that anyone can use to label AI content, with no attribution needed. There’s a basic icon, a “Fully AI-Generated” version and a “Partially AI-Modified” version, each in black, white and two 50% transparent variants, as SVG and PNG. The Commission is clear that using them is optional.
From the EU icons page (updated 24 September 2026):
- “Fully AI-Generated” is for content entirely generated by AI with no human-created elements apart from prompting. That’s an AI influencer’s photos.
- “Partially AI-Modified” is for human-made content partly changed with AI, like a real photo with an AI background.
- Placement: clearly perceivable at the latest at first exposure, and directly embedded into the content (the page makes an exception for creative works). The page says the icon “must be visible when content is reshared or downloaded”.
- The key line: “The use of these EU icons is optional, but the labelling requirements under Article 50 AI Act are not.”
What is the Code of Practice on AI labelling?
A voluntary rulebook the Commission published on 10 June 2026 for marking and labelling AI-generated content. Signing it is one way to show you comply with Article 50. It doesn’t give a legal presumption of conformity, and you don’t have to sign it. The Commission’s FAQ says it has been assessed as adequate.
It’s written for providers and deployers of generative AI (Commission announcement). The Digital Omnibus says codes like this “have limited legal effect, and in particular do not grant a presumption of conformity” (Regulation (EU) 2026/1744, recitals). Those who don’t sign must show compliance “through alternative adequate means”, per the FAQ. For a solo creator, that means the visible labels above, used consistently, and a note of what you label and where.
What happens on 2 December 2026?
Two things, neither of which delays your labelling. AI tools already on the market before 2 August 2026 get until 2 December 2026 to add machine-readable marks. And from the same day, the EU bans AI systems used to make realistic intimate images of identifiable people without their consent. Your duty as a deployer started on 2 August.
Both dates come from the Digital Omnibus on AI, Regulation (EU) 2026/1744 (adopted 8 July 2026, published 24 July 2026). Two more points from the Commission’s FAQ:
- No retroactive labels required. Content made before 2 August 2026 doesn’t need labelling after the fact, though the Commission encourages it.
- Don’t wait for your tools. A tool adding hidden marks in December doesn’t do your visible disclosure for you.
The intimate-images ban is about real, identifiable people. It’s one more reason to build only original characters: see AI likeness laws.
What are the fines for breaking Article 50?
Up to €15 million or 3% of total worldwide annual turnover for the previous year, whichever is higher, under Article 99(4) of the Act. For SMEs and start-ups, Article 99(6) caps it at whichever of the two is lower. National market surveillance authorities enforce it and weigh the size of the business and how serious the breach was.
Our view: for a one-person account the realistic risk is likely smaller than the headline, but it’s not zero, and platforms act faster than regulators. An unlabelled AI account can lose reach or content on Instagram and TikTok long before any authority writes to you.
What can go wrong?
- You treat her as a hobby. Once she earns on a regular basis, the Commission’s FAQ treats it as professional activity. Label like a deployer.
- You rely on metadata. The Commission says the machine-readable marks don’t fulfil your disclosure. Add a visible label.
- You crop the label out. A watermark or icon in a corner you later crop for a story is gone. Place it where every crop keeps it, or label the story itself.
- You wait for 2 December. That date is for tool makers’ hidden marks. Your duty started on 2 August 2026.
- You assume “fictional” means exempt. Fiction gets a lighter way to disclose, not an exemption.
- You follow this page as legal advice. It’s a practical reading of the texts linked above. For anything serious, ask a lawyer who knows EU law.
Questions people ask
Does the EU AI Act require labels on AI-generated images?
Does the EU AI Act apply to AI influencers?
Do I have to use the EU AI icon?
Does the EU AI Act apply if I live outside the EU?
What are the fines for not labelling under the EU AI Act?
Read next
- Are AI influencers legal? The 2026 rules, platform by platform
Instagram, Meta, TikTok, Fanvue, OnlyFans, the FTC, the EU AI Act and the TAKE IT DOWN Act: what each requires from an AI influencer. Checked 2 Oct 2026.
- Instagram's AI-generated profile label: who needs it and how to turn it on
Since 31 Aug 2026 Instagram limits AI personas without the AI-generated profile label. Who needs it, how to turn it on, and what it does to reach.
- How to add the AI label on TikTok (photos and videos)
The exact taps for TikTok's AI-generated content label on photo posts and videos, when TikTok labels you automatically, and whether it hurts reach.
- AI likeness laws: what's in force, what isn't, and how to stay clear
Which AI likeness laws are in force, which are still bills, and why an AI influencer should be an original character. Plus a lookalike self-check.